3

I am using the stepAIC function in R to do a bi-directional (forward and backward) stepwise regression. I do not understand what each return value from the function means.

The output is:

          Df     Sum of Sq    RSS       AIC
<none>                        350.71   -5406.0
- aaa      1     0.283        350.99   -5405.9
- bbb      1     0.339        351.05   -5405.4
- ccc      1     0.982        351.69   -5400.5
- ddd      1     0.989        351.70   -5400.5

Question Are the values listed under Df, Sum of Sq, RSS, and AIC the values for a model where only one variable would be considered as the independent variable (i.e. y ~ aaa, y ~ bbb, etc.)?

NOTE: Question also in data science exchange: data science link

1 Answers1

4

After consulting with someone I found out that the "none" category corresponds to a model that would include all the variables, in other words none of the variables were removed. So consider the line in the output for the variable aaa. The listed RSS and AIC are the values for a model that would include all variables but aaa and we see an increase in the RSS and AIC. The other listed results can be considered in the same fashion. The best model is then the one where none are removed since this has the smallest AIC.